Abstract: The spatial variability and the cause of the variability of soil arsenic (As) in north predominant agricultural area, Jiangsu Province were analyzed using mathematical statistics and geostatistic methods. Based on the spatial data analysis of the observed values, polynomial-weighted and cross-validation were adopted to evaluate the semivariogram model fitting results by spherical model, exponential model and Gaussian model. According to the evaluation results, the Gaussian model with sill was suggested to be preferable for fitting the semivariograms. Ordinary kriging were used to estimate soil As distribution in the monitoring area. The results show that soil As in north predominant agricultural area, Jiangsu Province, is autocorrelated and the experimental semivariogram is isotropic. The values of soil As are the highest in the north-west part of the monitoring area and the lowest values are in the middle-north part. Irrigation with poor quality water is the main cause of the variability.